Abstract

Introduction: Shrimp farming plays a crucial role in Vietnam's aquaculture industry, yet frequent disease outbreaks, primarily caused by poor water quality control, continue to present significant chal-lenges and economic losses. To address this issue, this study proposes an Artificial Intelligence (AI)- enhanced Internet of Things (IoT)-based intelligent monitoring system aimed at improving shrimp health and promoting sustainable farming practices. Methods: The proposed system consists of an IoT framework that continuously monitors key water quality parameters, including pH, temperature, salinity, and dissolved oxygen, using real-time sensor networks, while pond-mounted cameras periodically acquire high-resolution images of shrimp. An AI-driven diagnostic model based on a transfer-learned ResNet-50 architecture analyzes these images to detect and classify diseases, including black spot disease, black gill disease, and white spot syndrome virus (WSSV), fusing its predictions with threshold-based anomaly detection in multiparameter sensor data. Results: Several algorithms were trained on the proposed four-class shrimp disease dataset; of these, ResNet-50 demonstrated the optimal performance, achieving an accuracy of 0.8559, a precision of 0.8597, a recall of 0.8559, and an F1 score of 0.8552, while mitigating overfitting. In edge deployment on a Raspberry Pi 4, the system sustained an average image classification latency of approximately 120 ms and completed the end-to-end capture to alert pipeline in less than 300 ms during more than 1,000 inference runs. Conclusion: The results obtained indicate that integrating AI with IoT technologies holds considerable potential for preventing shrimp disease outbreaks, reducing economic losses, and fostering sustainable aquaculture practices in Vietnam and similar intensive shrimp farming regions.